Transarterial radioembolization vs transarterial chemoembolization with drug-eluting beads for treating hepatocellular carcinoma: a cost-effectiveness analysis in Japanese healthcare system

We developed a Markov model to compare the cost-effectiveness of TARE and DEB-TACE for HCC. Markov processes can simulate the long-term health states of a patient by tracking the patient's transitions between the health states defined in the model. During each cycle, patients transition according to the transition probabilities and accumulate the quality-adjusted life year (QALY) and cost for each state. Our model spans 5 years of 2-month cycles based on previous studies [6, 14, 15], with five health states: local, transarterial embolization (TAE)-eligible progressive, TAE-refractory progressive, decompensated cirrhosis, and death (Fig. 1). Our hypothetical cohort included patients with unresectable HCC who were not eligible for percutaneous ablation, partial hepatectomy, or transplantation based on the inclusion criteria set by an RCT comparing TARE and DEB-TACE [6]. Patients began in the local state and received TARE or DEB-TACE within the first cycle. Subsequently, patients remained in the local state, transitioned to the TAE-eligible progressive state, developed non-compensated cirrhosis, or died based on transition probabilities. Patients in the TAE-eligible progressive state were still eligible for TARE or DEB-TACE. Patients who progressed further after two TARE or DEB-TACE sessions were ineligible for additional sessions and transitioned to a TAE-refractory progressive state according to the definition of “TACE failure” in the Japanese Clinical Practice Guidelines for HCC [5]. Patients in the TAE-refractory progressive state, following the algorithm for drug therapy in the Japanese guideline, were recommended to receive treatment regimens, including immune checkpoint inhibitors (atezolizumab + bevacizumab). In recent years, in actual clinical practice, systemic therapy is often started before chemoembolization, and then, chemoembolization is performed on demand. However, in this study, to simplify the model and to conform to the Japanese guideline, we used a model in which chemoembolization was performed first, and systemic therapy was performed only in patients who became TAE refractory. The total maximum sessions for DEB-TACE and TARE were set at six and two, respectively, based on previous studies [6, 14]. Patients in other health states transitioned to decompensated cirrhosis, indicating the need for palliative care. Costs and outcomes were discounted at 2% annually, following the Japanese guideline for economic evaluation of drugs and medical devices [22].

Transition probabilities were calculated using survival and decompensation rate data from previous studies [6, 23‐26]. In the primary analysis, we used survival data from the intention-to-treat analysis of an RCT that compared TARE and DEB-TACE [6] to calculate transition probabilities, whereas we used survival data from the per-protocol analysis in the ancillary analysis. Transition probabilities were calculated using a method similar to a previous cost-effectiveness analysis of small-cell lung cancer treatment [27]. Based on the assumption that the survival curves of previous trials can be approximated by an exponential function, the transition probabilities were calculated using the following formula based on the previous study, which provides the equations: P (transition probabilities) = 1 − (0.5)^{(length of cycle/median time to event)} [27]. Our model also considers the development of decompensated cirrhosis with certain transition probabilities based on previous cohort studies [24, 25]. As shown in Fig. 1, the transition probability from local to local equals that of TAE-eligible progressive to TAE-eligible progressive (P1 = P1'); transition probability from local to TAE-eligible progressive equals that of TAE-eligible progressive to TAE-refractory progressive (P2 = P2'); the transition probability from local to decompensated cirrhosis equals that of TAE-eligible progressive to decompensated cirrhosis (P3 = P3'), and the transition probability from local to death equals that of TAE-eligible progressive to death (P4 = P4') based on the assumption that survival curves that approximate the same exponential function are applied in states eligible for TARE or DEB-TACE. Appendix E1 provides detailed calculation formulas.

The utility data were estimated from quality-of-life scores based on the EuroQol 5 dimensions scale from a previous study on HCC and hepatitis C virus-related diseases in Japan [28] and another study evaluating the cost-effectiveness of lenvatinib versus sorafenib for the treatment of patients with unresectable HCC in Japan using data from the REFLECT Trial [29]. The quality-of-life scores of the “local state” and “TAE-refractory progressive state” were derived from the mean of “HCC-intermediate” [28] and “progression-free” [29], and the mean of “HCC-advanced” [28] and”progressed” [29], respectively, reported by previous studies. The quality-of-life score of the “TAE-eligible progressive state” is estimated based on the assumption that it is the mean of the score of the “local state” and “TAE-refractory progressive state.”

The costs of DEB-TACE were calculated using the DPC database [21], a Japanese nationwide inpatient database that includes hospital administrative claims data and discharge abstracts for 8 million inpatients in over 1,200 hospitals throughout Japan. It includes data recorded during hospitalization (sex, age, diagnosis, surgical procedure, International Classification of Diseases 10th Revision [ICD-10] codes, and total hospitalization cost). This database was used to identify patients with HCC treated using DEB-TACE between April 2018 and March 2022. The following search criteria were used to select patients from the database: (1) claiming selective TACE (Japanese procedure claiming code; K615_2) as the procedure fee; (2) HCC (ICD-10 code: C220) was the major disease requiring hospitalization; (3) claiming drug-eluting beads; and (4) not claiming lipiodol (excluding conventional TACE). Hospitalization costs for the following two patient cohorts were considered: patients solely reimbursed for TACE (without severe hepatobiliary complications) and those with severe hepatobiliary complications. Patients with severe hepatobiliary complications were defined as those who claimed reimbursement for the following procedures after DEB-TACE: percutaneous gallbladder or liver abscess drainage (claim code: J010-2), surgical biliary drainage (claim codes: K682, K696, K697), percutaneous transhepatic biliary drainage (claim code: K682-2), endoscopic biliary drainage (claim codes: K682-3), endoscopic biliary stent (claim code: K688), percutaneous transhepatic biliary stent (claim code: K689), surgical liver abscess drainage (claim code: K691), percutaneous liver abscess drainage (claim code: K691-2), and surgical liver abscess resection (claim code: K694). The number of patients with severe hepatobiliary complications and the cost of hospitalization for these patients were extracted from the DPC database. The cost of potential complications considering the occurrence of severe hepatobiliary complications was estimated. The weighted average of these costs represented the average hospitalization cost for DEB-TACE applied to the local and TAE-eligible progressive states.

The costs of TARE were estimated differently because TARE lacks approval from Japanese health insurance companies. The costs of DEB-TACE were categorized into components assuming that the shared costs between TARE and DEB-TACE, such as procedure fees, reimbursable devices (sheath, catheter, microcatheter, and guidewire), daily hospitalization costs, and potential complications, would be identical. In some countries, TARE is performed as an outpatient procedure; however, we assumed that it would be performed as an admission procedure in Japan. Costs unique to TARE, such as pre-procedure angiography, scintigraphy, and cost of radioactive microspheres, were considered additional expenses. The costs of pre-procedural angiography and scintigraphy were added only once for the first TARE session. The costs of pre-procedure angiography, scintigraphy, and contrast computed tomography for follow-up in the local and TAE-eligible progressive states were determined based on the national fee schedule. However, the cost of radioactive microspheres was derived from a study conducted in the United Kingdom [11]. To estimate the potential costs for early severe complications, we assumed that there was no significant difference in the risk of severe complications between TARE and DEB-TACE according to the results of a previous RCT that found no significant differences in the frequency of participants with at least one serious adverse event until 6 months after treatment and thirty-day mortality between the TARE and DEB-TACE arms [6].

The costs of treatment regimens involving immune checkpoint inhibitors (atezolizumab + bevacizumab) for the TAE-refractory progressive state were sourced from the White Paper published by the Japan Society of Hepatology [30]. Costs for decompensated cirrhosis and outpatient therapy for the local and TAE-eligible progressive states were extrapolated from previous Japanese studies using real-world claim data for viral hepatitis-related diseases and nonalcoholic fatty liver disease [31‐33].

The current study adhered to the Consolidated Health Economic Evaluation Reporting Standards 2022 Statement and Japanese guidelines [22, 34, 35]. The perspective considered was that of healthcare payers under the Japanese system, including only the monetary costs directly attributed to disease management and excluding indirect costs such as productivity loss caused by the disease [36]. The incremental cost-effectiveness ratio (ICER) was calculated as the incremental cost per QALY of TARE compared with that of DEB-TACE and compared with a benchmark for the willingness-to-pay [5 million Japanese yen (JPY) per QALY [37]](31,250 USD/QALY). The ICER must be smaller than the willingness-to-pay threshold for the treatment to be considered cost-effective. In other words, TARE would be considered socially acceptable if the incremental cost to gain one incremental QALY compared to DEB-TACE is within the willingness-to-pay threshold (5 million JPY). The costs and ICER are also expressed in US dollars using a currency exchange rate of JPY 160 per 1 US dollar.

One-way deterministic sensitivity analysis was conducted to identify input parameters that exerted the most substantial influence on ICER. Cost and utility inputs were subjected to a deviation of 10%, whereas survival data employed in generating transition probabilities were subjected to a deviation of 20–30% according to the original research’s standard deviation. The discount rate range was determined as 0–4% according to the Japanese guideline [22].

Furthermore, probabilistic sensitivity analysis (Monte Carlo simulation) was performed to simulate 10,000 patients in the TARE and DEB-TACE cohorts. Based on previous research, probability and utility inputs followed beta distributions with a standardized difference of 10%, whereas cost inputs followed gamma distributions with a similar standardized difference [11]. The results of these simulations generated acceptability curves, illustrating the probability of cost-effectiveness at different willingness-to-pay thresholds.

Analyses were conducted using Stata/MP 16.0 (StataCorp, College Station, TX, USA) and TreeAge Pro Healthcare 2023 (TreeAge Software, Inc., Williamstown, MA, USA).

Tables 1 and 2 list the baseline parameters, including the utilities and costs in the model and transition probabilities, respectively. Figure 2 shows the projected survival curves based on these estimated probabilities.

From the DPC database, we identified 6,986 patients (74.86 ± 9.67 years old; 5196 males, 1,790 females; 6961 without any complication requiring interventional procedures, 25 patients with severe hepatobiliary complications requiring interventional procedures) with HCC treated with DEB-TACE between April 2018 and March 2022. The age of patients identified from the DPC database was comparable to the mean age at clinical diagnosis of HCC in Japan (70.0 and 74.0 years for men and women, respectively [38]) and tended to be higher than that of patients in the RCT used to calculate the transition probability (median age: 67 years for the TARE arm and 68 years for the DEB-TACE arm [6]). Based on the DPC data, the mean cost of hospitalization for DEB-TACE was 839,223 JPY (5245 USD) without complications and 2,543,046 JPY (15,894 USD) with severe hepatobiliary complications. The cost of potential complications considering the occurrence of severe hepatobiliary complications was 6097 JPY (38 USD). Based on these data, the weighted average cost of DEB-TACE was estimated as 845,320 JPY (5,283 USD).

Using these data, the cost of hospitalization for TARE (costs other than radioactive microspheres) was estimated to be 734,896 JPY (4,593 USD). The cost of radioactive microspheres was calculated as 1,440,000 JPY (9000 USD)[11], and the total cost of hospitalization for TARE was estimated as 2,174,896 JPY (13,593 USD). Table 3 lists the detailed cost estimations for TARE. These data were input into the model as the costs of the local and TAE-eligible progressive states.

Table 1 lists the cost and QALY of each therapy based on the Markov model. In the primary analysis, using the survival data from the intention-to-treat analysis result, the 5-year follow-up costs were estimated to be 7,499,735 JPY (46,873 USD) and 9,853,942 JPY (61,587 USD) for DEB-TACE and TARE, respectively. The incremental QALYs were estimated as 1.22 years and 1.68 years for DEB-TACE and TARE, respectively. The ICER was 5,173,591 JPY/QALY (32,335 USD/QALY), which exceeded the benchmark of willingness-to-pay (5 million JPY/QALY [37]). Using the survival data of the per-protocol analysis result in the ancillary analysis, the ICER was 4,156,533 JPY/QALY (25,978 USD/QALY), below the willingness-to-pay benchmark. These data are summarized in Table 4.

Figure 3 shows the results of the one-way deterministic sensitivity analysis (performed in the primary analysis) as a tornado diagram. The parameter with the greatest influence on the ICER was the median PFS, which was used to calculate the transition probabilities from the local state to the TAE-eligible progressive state and from the TAE-eligible progressive state to the TAE-refractory progressive state. The one-way deterministic sensitivity analysis in the primary analysis suggested that reducing the reimbursement price of radioactive microspheres from 1.440 million JPY (9,000 USD) to 1.399 million JPY (8,745 USD), approximately 2.8% lower than the price in the United Kingdom, would align the ICER with the 5 million JPY/QALY threshold. In contrast, the impact of the utility of each state on the ICER was small.

In the primary analysis, using the survival data of the intention-to-treat analysis result, less than half of the estimates in the probabilistic sensitivity analysis with Monte Carlo simulations were located below the willingness-to-pay benchmark of 5 million JPY/QALY (Fig. 4). Therefore, TARE was less likely to be cost-effective than DEB-TACE in the primary analysis. Figure 5 shows the acceptability curve based on probabilistic sensitivity analysis. In the primary analysis, when the willingness-to-pay was approximately more than 5.17 million JPY (approximately 32,300 USD), TARE was likely to be more cost-effective than DEB-TACE.